Contour Computation of Linearized Painlevé II and IV Solutions with Monodromy-Based Error Control

Abstract

We study the numerical evaluation of contour integral representations for solutions of the linearized second and fourth Painlevé equations. The construction combines a nonlinear background solution, continuation of canonical Lax-pair columns, normalization matching, and quadrature over cycles with decaying branches in distinct sectors. Accuracy is assessed using a reference fundamental matrix, differential-equation residuals, and variations of monodromy data computed independently from the spectral problem. An exact identity describes the propagation of initial-basis errors. A stagewise a posteriori refinement criterion uses increments in monodromy variations to allocate numerical accuracy. The contour algorithm is compared with the Dormand--Prince method on a regular background, in a region of rapid solution growth, and through hard loss of stability followed by approximately regular oscillations. Accumulated offsets in monodromy variations and subsequent drift are measured separately. The experiments show how drift depends on spectral conditioning, quadrature accuracy, and error allocation between initial and subsequent parts of the computation. Five additional initial-data and parameter cases demonstrate the dependence of comparative performance on the nonlinear background.